From patchwork Thu Dec 2 07:40:34 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Wenbin Chen X-Patchwork-Id: 31875 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a6b:d206:0:0:0:0:0 with SMTP id q6csp64040iob; Wed, 1 Dec 2021 23:45:05 -0800 (PST) X-Google-Smtp-Source: ABdhPJwnVVWwfmeVwYHfdUjh/dEVOFkYBXB50uC4jz1tbygM3Ftvo8PwP5jx6Mzp8txu4//P5CAr X-Received: by 2002:aa7:dc14:: with SMTP id b20mr15402055edu.133.1638431105045; Wed, 01 Dec 2021 23:45:05 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1638431105; cv=none; d=google.com; s=arc-20160816; b=DyD3hSWlRe1duUN77g0Qeo5tOabR5efD2Rfw/pSxN7pUJSA2ffyVU74D//FRpq/2UB wyOVHMFJDp4UdsZNlnIC5g+FYYTuiuFkG2KntD/AvGFndkR8FvjUGr7VPy/t078Jkanb ltGxLAJ+qC/7r8in7aLYqrn5zBRDG5RP+NXsLspbTh2SBd9+dNuhEzgJAWFcSLZqSX0b R9paCWWNpDyk2xyQONvsVOjC3kPPELzq51V5mNHu+GKPeFnGt4AhRp13SyVxi/U/Be0H Y9J874/QlFdreXunvLGdYSXZ6Jm1CnULhwB9KJlwdcZuo6xO9E98FHrioDZnUpA3viiC FIdA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:cc:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:mime-version:message-id:date:to:from :delivered-to; bh=rVzPjbIecXDD9t8jg2XUuMRbAbjLhsOsyU1d4ybr//8=; b=r4ejVsyqeIx5nJe0Xz5zZ0bzsSQF3HHZOBghb4qvW9TTMu7THZy9Ih99yJT3kDxRsG rjVc7rS3XCl7uPRpfFWNuM/NOhIG6/L9RFhLelsSCXfDEmaK7K2KlMCcFHDSUjyFbfzB wCDjfzuCFV6jwuoM+PxkIFGJb99R/i9plg89nUrkVrjZWklOzpuVcWfP411EOpAFoyHr T8O7nLOX/6xZBWDqdcWgRAxR0GV8hiiluHwmClpYkiazzJ4NqTUPWQMcj6AXO+61BSMn TzSXt3CUx+oVUuoKsIOxR4xV0DOiJIe7GVupu6m8udadHKJ78J2Sj0oHwj5FyPPlgQ1h /I5w== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=intel.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id l18si3829134edb.148.2021.12.01.23.45.04; Wed, 01 Dec 2021 23:45:05 -0800 (PST) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 792AD68AE5D; Thu, 2 Dec 2021 09:45:00 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mga11.intel.com (mga11.intel.com [192.55.52.93]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 155AB68AA83 for ; Thu, 2 Dec 2021 09:44:52 +0200 (EET) X-IronPort-AV: E=McAfee;i="6200,9189,10185"; a="234157406" X-IronPort-AV: E=Sophos;i="5.87,281,1631602800"; d="scan'208";a="234157406" Received: from orsmga008.jf.intel.com ([10.7.209.65]) by fmsmga102.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 01 Dec 2021 23:44:50 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.87,281,1631602800"; d="scan'208";a="513078941" Received: from chenwenbin-z390-aorus-ultra.sh.intel.com ([10.239.35.110]) by orsmga008.jf.intel.com with ESMTP; 01 Dec 2021 23:44:49 -0800 From: Wenbin Chen To: ffmpeg-devel@ffmpeg.org Date: Thu, 2 Dec 2021 15:40:34 +0800 Message-Id: <20211202074034.1163615-1-wenbin.chen@intel.com> X-Mailer: git-send-email 2.25.1 MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH] libavutil/hwcontext_qsv: clean padding when upload qsv frames X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Cc: Wenbin Chen Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: JVaegmABfwBp When we upload a frame that is not padded as MSDK requires, we create a new AVFrame to copy data. The frame's padding data is uninitialized so it brings run to run problem. For example, If we run the following command serveral times we will get different outputs. ffmpeg -init_hw_device qsv=qsv:hw -qsv_device /dev/dri/renderD128 -filter_hw_device qsv -f rawvideo -s 192x200 -pix_fmt p010 -i 192x200_P010.yuv -vf "format=nv12,hwupload=extra_hw_frames=16" -c:v hevc_qsv output.265 According to https://github.com/Intel-Media-SDK/MediaSDK/blob/master/doc/mediasdk-man.md#encoding-procedures "Note: It is the application's responsibility to fill pixels outside of crop window when it is smaller than frame to be encoded. Especially in cases when crops are not aligned to minimum coding block size (16 for AVC, 8 for HEVC and VP9)" I add a function to fill padding area with border pixel to fix this run2run problem, and also move the new AVFrame to global structure to reduce redundant allocation operation to increase preformance. Signed-off-by: Wenbin Chen --- libavutil/hwcontext_qsv.c | 96 +++++++++++++++++++++++++++++++++------ 1 file changed, 83 insertions(+), 13 deletions(-) diff --git a/libavutil/hwcontext_qsv.c b/libavutil/hwcontext_qsv.c index 268be9f8a1..983494666b 100644 --- a/libavutil/hwcontext_qsv.c +++ b/libavutil/hwcontext_qsv.c @@ -47,6 +47,7 @@ #include "pixfmt.h" #include "pixdesc.h" #include "time.h" +#include "imgutils.h" #define QSV_VERSION_ATLEAST(MAJOR, MINOR) \ (MFX_VERSION_MAJOR > (MAJOR) || \ @@ -90,6 +91,7 @@ typedef struct QSVFramesContext { mfxExtOpaqueSurfaceAlloc opaque_alloc; mfxExtBuffer *ext_buffers[1]; + AVFrame realigned_tmp_frame; } QSVFramesContext; static const struct { @@ -137,6 +139,54 @@ static uint32_t qsv_get_d3d11va_bind_flags(int mem_type) } #endif +static int qsv_fill_border(AVFrame *dst, const AVFrame *src) +{ + const AVPixFmtDescriptor *desc; + int i, planes_nb = 0; + if (dst->format != src->format) + return AVERROR(EINVAL); + + desc = av_pix_fmt_desc_get(dst->format); + + for (i = 0; i < desc->nb_components; i++) + planes_nb = FFMAX(planes_nb, desc->comp[i].plane + 1); + + for (i = 0; i < planes_nb; i++) { + int sheight, dheight, y; + ptrdiff_t swidth = av_image_get_linesize(src->format, + src->width, + i); + ptrdiff_t dwidth = av_image_get_linesize(dst->format, + dst->width, + i); + const AVComponentDescriptor comp = desc->comp[i]; + if (swidth < 0 || dwidth < 0) { + av_log(NULL, AV_LOG_ERROR, "av_image_get_linesize failed\n"); + return AVERROR(EINVAL); + } + sheight = src->height; + dheight = dst->height; + if (i) { + sheight = AV_CEIL_RSHIFT(src->height, desc->log2_chroma_h); + dheight = AV_CEIL_RSHIFT(dst->height, desc->log2_chroma_h); + } + //fill right padding + for (y = 0; y < sheight; y++) { + void *line_ptr = dst->data[i] + y*dst->linesize[i] + swidth; + av_memcpy_backptr(line_ptr, + comp.depth > 8 ? 2 : 1, + dwidth - swidth); + } + //fill bottom padding + for (y = sheight; y < dheight; y++) { + memcpy(dst->data[i]+y*dst->linesize[i], + dst->data[i]+(sheight-1)*dst->linesize[i], + dwidth); + } + } + return 0; +} + static int qsv_device_init(AVHWDeviceContext *ctx) { AVQSVDeviceContext *hwctx = ctx->hwctx; @@ -220,6 +270,7 @@ static void qsv_frames_uninit(AVHWFramesContext *ctx) av_freep(&s->surface_ptrs); av_freep(&s->surfaces_internal); av_freep(&s->handle_pairs_internal); + av_frame_unref(&s->realigned_tmp_frame); av_buffer_unref(&s->child_frames_ref); } @@ -1014,12 +1065,13 @@ static int qsv_transfer_data_to(AVHWFramesContext *ctx, AVFrame *dst, QSVFramesContext *s = ctx->internal->priv; mfxFrameSurface1 in = {{ 0 }}; mfxFrameSurface1 *out = (mfxFrameSurface1*)dst->data[3]; + mfxFrameInfo tmp_info; mfxSyncPoint sync = NULL; mfxStatus err; int ret = 0; /* make a copy if the input is not padded as libmfx requires */ - AVFrame tmp_frame; + AVFrame *tmp_frame = &s->realigned_tmp_frame; const AVFrame *src_frame; int realigned = 0; @@ -1048,24 +1100,40 @@ static int qsv_transfer_data_to(AVHWFramesContext *ctx, AVFrame *dst, if (ret < 0) return ret; + /* According to MSDK spec for mfxframeinfo, "Width must be a multiple of 16. + * Height must be a multiple of 16 for progressive frame sequence and a + * multiple of 32 otherwise.", so allign all frames to 16 before uploading. */ if (src->height & 15 || src->linesize[0] & 15) { realigned = 1; - memset(&tmp_frame, 0, sizeof(tmp_frame)); - tmp_frame.format = src->format; - tmp_frame.width = FFALIGN(src->width, 16); - tmp_frame.height = FFALIGN(src->height, 16); - ret = av_frame_get_buffer(&tmp_frame, 0); - if (ret < 0) + if (tmp_frame->format != src->format || + tmp_frame->width != FFALIGN(src->width, 16) || + tmp_frame->height != FFALIGN(src->height, 16)) { + av_frame_unref(tmp_frame); + + tmp_frame->format = src->format; + tmp_frame->width = FFALIGN(src->width, 16); + tmp_frame->height = FFALIGN(src->height, 16); + ret = av_frame_get_buffer(tmp_frame, 0); + if (ret < 0) + return ret; + } + ret = av_frame_copy(tmp_frame, src); + if (ret < 0) { + av_frame_unref(tmp_frame); return ret; - - ret = av_frame_copy(&tmp_frame, src); + } + ret = qsv_fill_border(tmp_frame, src); if (ret < 0) { - av_frame_unref(&tmp_frame); + av_frame_unref(tmp_frame); return ret; } + + tmp_info = out->Info; + out->Info.CropW = FFMIN(out->Info.Width, tmp_frame->width); + out->Info.CropH = FFMIN(out->Info.Height, tmp_frame->height); } - src_frame = realigned ? &tmp_frame : src; + src_frame = realigned ? tmp_frame : src; if (!s->session_upload) { if (s->child_frames_ref) @@ -1097,8 +1165,10 @@ static int qsv_transfer_data_to(AVHWFramesContext *ctx, AVFrame *dst, return AVERROR_UNKNOWN; } - if (realigned) - av_frame_unref(&tmp_frame); + if (realigned) { + out->Info.CropW = tmp_info.CropW; + out->Info.CropH = tmp_info.CropH; + } return 0; }